agentera — plugin by jgabor

Agentera is a skill suite for autonomous software development, enhancing project management and decision-making.

<strong>Skill suite</strong> for autonomous software development.

</div>

Install

npx skills install -g jgabor/agentera

Installs all 12 skills for any supported runtime: Claude Code, OpenCode, Cursor, Codex, Gemini CLI, and 40+ others. Type /hej to begin.

Lifecycle hooks (optional)

Hooks add session context preload, artifact validation, and session bookmarks. Without hooks, skills work independently; with hooks, they behave as a team.

Claude Code: hooks auto-load from the installed skill directory. No extra step.

OpenCode: copy the hook plugin:

curl -fsSL https://raw.githubusercontent.com/jgabor/agentera/main/.opencode/plugins/agentera.js \
  -o ~/.config/opencode/plugins/agentera.js

Alternative install methods

Claude Code plugin registry: claude plugin add jgabor/agentera

Manual (git clone):

git clone [email protected]:jgabor/agentera.git ~/.agents/agentera

Then reference globally in your runtime's settings (Claude Code example):

{
  "skillPaths": ["~/.agents/agentera/skills"]
}

[!NOTE] profilera mines runtime-specific session data and remains adapter-specific per the Section 21 Session Corpus Contract. All other skills are fully portable. For OpenCode capability details, see references/adapters/opencode.md.

Getting started

Type /hej and agentera reads your entire project (code, git history, open issues, health grades) and tells you where things stand:

┌─┐┌─┐┌─┐┌┐┌┌┬┐┌─┐┬─┐┌─┐
├─┤│ ┬├┤ │││ │ ├┤ ├┬┘├─┤
┴ ┴└─┘└─┘┘└┘ ┴ └─┘┴└─┴ ┴

─── status ─────────────────────────────

  ⛶ health    ⮉ B+ (testing: C)
  ⇶ issues    0 critical · 2 degraded · 5 annoying
  ≡ plan      [██████▓░░░] 6/10 tasks
  ♾ profile   loaded

  Shipped auth middleware and rate limiting last cycle.
  Health trending up, test coverage still lagging.

─── attention ──────────────────────────

  ⇉ test coverage below 60%, degrading since cycle 8
  ⇉ task 7 blocked on API schema decision

─── next ───────────────────────────────

  suggested → ❈ /resonera (resolve API schema to unblock task 7)

How it works

Skills communicate through markdown files in your project: a vision doc, a plan, a health report, a decision log. Each skill reads what the others have written and acts on it. You don't manage these files; they build up naturally as you work.

Every skill suggests what to do next when it finishes. You follow the thread, or run /orkestrera to execute an entire plan: it dispatches skills, evaluates each task with inspektera, retries failures, and loops until done.

Example workflow

Continuing from the /hej output above (task 7 blocked on an API schema decision):

/resonera                                         decide
├─ Deliberates on the API schema tradeoffs
├─ Writes Decision 8 → DECISIONS.md
└─ Suggests → /planera

/planera                                           plan
├─ Reads the decision, breaks remaining work into 4 tasks
├─ Each task gets behavioral acceptance criteria
└─ Suggests → /orkestrera

/orkestrera                                        execute
├─ Task 7  → realisera builds → inspektera evaluates → PASS
├─ Task 8  → realisera builds → inspektera evaluates → PASS
├─ Task 9  → realisera builds → inspektera evaluates → FAIL → retry → PASS
├─ Task 10 → realisera builds → inspektera evaluates → PASS
├─ Plan complete. Runs full health audit.
└─ Health: B+ → A-

Each skill writes markdown artifacts (a vision, a plan, a health report, a decision log). The next skill reads what the last one wrote and acts on it. You don't manage these files; they build up naturally as you work. profilera watches how you make decisions and tunes every skill to your preferences over time.

Skills

	Skill	What it does
⌂	hej	Entry point. Reads your project state, shows what needs attention, suggests where to start.
⛥	visionera	Envision. Defines and evolves your project's north star through codebase exploration and aspirational challenge.
❈	resonera	Deliberate. Thinks through hard decisions via Socratic questioning before you commit.
⬚	inspirera	Research. Analyzes an external resource and maps its patterns to your project.
≡	planera	Plan. Breaks work into tasks with clear done-criteria, scales from quick notes to full plans.
⧉	realisera	Build. Autonomous development loop that picks up work, implements, verifies, and continues.
⎘	optimera	Tune. Picks a metric, runs experiments, measures results, iterates until it improves.
⛶	inspektera	Audit. Audits code health across nine dimensions (architecture, patterns, coupling, complexity, tests, deps, versioning, artifact freshness, security), tracks trends over time.
▤	dokumentera	Document. Creates and maintains docs, tracks what's covered and what's missing.
♾	profilera	Compounding memory. Mines your decision patterns into a profile consumed by every skill, so the 20th cycle adapts to how you work in ways the 1st could not.
◰	visualisera	Visualize. Creates and maintains a visual identity system for your project.
⎈	orkestrera	Orchestrate. Dispatches skills as subagents, evaluates each with inspektera, loops through plans.

<details> <summary><strong>State artifacts reference</strong></summary> <br>

Three project-facing files at root, nine operational files in .agentera/.

Root (project-facing):

Artifact	Maintained by	Consumed by
`VISION.md`	visionera, realisera	realisera, planera, inspektera, dokumentera, visualisera, orkestrera
`TODO.md`	realisera, inspektera	realisera, planera, orkestrera
`CHANGELOG.md`	realisera	project contributors

.agentera/ (operational):

Artifact	Maintained by	Consumed by
`PROGRESS.md`	realisera	planera, inspektera, dokumentera, visionera, orkestrera
`DECISIONS.md`	resonera	planera, realisera, optimera, inspektera, profilera, orkestrera
`PLAN.md`	planera	realisera, inspektera, orkestrera
`HEALTH.md`	inspektera	realisera, planera, orkestrera
`OBJECTIVE.md`	optimera	optimera
`EXPERIMENTS.md`	optimera	optimera
`DESIGN.md`	visualisera	realisera, visionera
`DOCS.md`	dokumentera	all skills (path overrides)
`SESSION.md`	session stop hook	session start hook, hej

PROFILE.md is global. In Claude Code, the reference implementation stores it at ~/.claude/profile/PROFILE.md. Other runtimes provide their own equivalent profile path through the host adapter contract.

</details>